General Report on Tunny


23F Page 89

23F     MESSAGE SLIDES


(a) Definition of Message Slides.

In statistical setting a few wrong letters of cipher do not matter much, but a single omitted letter or inserted letter makes it impossible to find any setting for the Χ's which is correct for the whole message. The effect of one (or more) omitted or inserted letters is called a Message Slide. It does not necessarily make it impossible to find settings, for, with the Χ's set correctly for one part of the message. ΔD will have systematic bulges on this part, which will not be greatly changed by the addition of the merely random bulges on the rest of the message: the sigma-ages will of course be reduced, because the text length is that of the whole message, the score that of a part only.


(b) Rival Settings.

A message slide can sometime be detected by inspection of the 1+2 scores. If several (say 3) letters are omitted, the settings of all Χ-wheels will be increased by three.
  A pair of scores such as    26 17   1398   5.9σ
    29 20   1253   4.1σ

suggests very strongly that there is a slide of 3 at a place dividing the tape roughly in the ratio 3:2; but see Antislides [23 G(d)].


(c) Spanning.

In any case, as soon as a message is set on any wheels, the relevant score is spanned, in thousands or in thirds whichever is the less. If some parts show no bulge there is probably a slide: this will increase the sigma-age on the remainder; it is worth while to span scores down to 3.7σ unless a certain setting already been obtained. Subsequent runs are done on a slide-free portion.


(d) Slide Runs.

When the parts of a message at slide settings are of considerable length, it is usual, after setting all wheels or even before to set the other parts by the means of a Slide Run. The Χ's are all set back a few places and are stepped together (upper switches on Colossus) counting those letters which are most numerous in the part already set, e.g. /U58, spanning a part not set. Because this is a short run (usually 50 positions at most) it is not necessary


< previous

next >